Agent-based modeling of AI-driven labor market transitions. Seven papers and 45,400+ Monte Carlo simulations exploring how universal basic income, monetary regimes, endogenous technology dynamics, household heterogeneity, and supply chain coupling interact to produce phase transitions in automation adoption.
A stock-flow consistent agent-based model (SFC-ABM) with 10,000 heterogeneous firms and 100,000 individual households across 6 sectors, calibrated to the Polish economy (GUS 2024). The series progressively relaxes assumptions — from static parameters to endogenous technology, dynamic networks, and heterogeneous household agents — testing whether the core finding (a reentrant phase transition at BDP ~500 PLN) survives each extension, and what it hides.
- Agent-based modeling with stock-flow consistent balance-sheet accounting
- Phase transitions & critical phenomena — bifurcation diagrams, susceptibility peaks, critical exponents
- Finite-size scaling and data collapse for universality class identification
- Network science — Watts-Strogatz, Erdos-Renyi, Barabasi-Albert topologies + endogenous rewiring
- Empirical estimation — GMM and hierarchical Bayesian (PyMC) on OECD panel data
- Input-output analysis — Leontief technical coefficient matrix calibrated from GUS symmetric I-O tables
- Factorial experimental design for mechanism isolation
- Heterogeneous households — 100,000 individual agents with savings, debt, skill decay, and health scarring
- Monte Carlo robustness — 30–100 seeds per parameter point across multi-dimensional sweeps
| # | Repo | Title | Sims | DOI |
|---|---|---|---|---|
| 1 | paper-01-acceleration-paradox |
The Acceleration Paradox | 6,300 |  |
| 2 | paper-02-monetary-regimes |
PLN vs EUR with SGP Constraint | 1,260 |  |
| 3 | paper-03-empirical-sigma |
Empirical CES σ Estimation (OECD, GMM + Bayesian) | 120 |  |
| 4 | paper-04-phase-diagram |
Phase Diagram & Universality | 18,540 |  |
| 5 | paper-05-endogenous |
Endogenous Technology & Network Dynamics | 10,080 |  |
| 6 | paper-06-heterogeneous-households |
Heterogeneous Households & Limits of UBI | 1,500 | pending |
| 7 | paper-07-io-coupling |
SFC-IO Sectoral Coupling: Three Propagation Channels | 3,630 | pending |
Engine: core — reusable Scala 3 SFC-ABM engine (sbt)
- Acceleration paradox: moderate UBI causes automation rather than responding to it (Paper 1)
- Monetary sovereignty matters: PLN float permits the transition; EUR + SGP kills it (Paper 2)
- σ calibration doesn't: 5–9× change in elasticity shifts adoption by only 1.5 pp (Paper 3)
- Topology universality: BDP_c = 500 PLN across all four network topologies, mean-field γ ≈ 1.0 (Paper 4)
- Endogenization preserves universality: reentrant shape survives; BDP_c shifts by at most 250 PLN (Paper 5)
- Aggregate metrics mask destruction: BDP_c = 500 appears as a "sweet spot" in aggregates but is the point of peak bankruptcy (17.3%), peak poverty (45%), and peak income Gini (0.50) at the household level (Paper 6)
- Scarring catch-22: unemployment erodes skills and health, making retraining least effective for those who need it most — even doubled intensity yields only 18% success (Paper 6)
- Supply chains are transformative: I-O coupling at κ=0.50 boosts adoption from ~12% to ~43% (+30 pp); bell-shaped response with NPL-driven reversal at κ ≈ 0.75 (Paper 7)
All repositories are released under the MIT License.